YOLOv11-ND: A Method for Identifying Traffic Targets in Nighttime Urban Environments

Traffic target recognition is a crucial technology that has drawn a lot of interest due to the quick development of unmanned and assisted driving systems. However, the precision and performance of target recognition for the more complicated nighttime environment are lower, and the majority of the present research on traffic target recognition concentrates on the daytime. By using nighttime traffic targets as the research object, this paper suggests YOLOv11-ND, an enhanced target recognition method, to address the aforementioned issues. First, based on WTConv, the WTC3k2 module is intended to take the position of C3k2 in the backbone part. This reduces the amount of model parameters without sacrificing precision. Then, to boost the fusion ability of multi-scale features, the HS-FPN network structure is adopted in the neck section. This improves the detection performance. Lastly, the model is optimized using the Focaler-GIoU loss function to further enhance the detection performance. In comparison to the baseline model YOLOv11, the enhanced model YOLOv11-ND improved the P, R, mAP50, and mAP50-95 measures by 3.1%, 3.4%, 4%, and 3.6%, respectively, according to experimental validation using the FLIR dataset.The method can successfully increase the precision of traffic target detection in urban settings at night, according to the testing results.

Abstract licence: CC BY-NC-ND

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.